Updated:
22 Apr 2008
Designing With Opamps. |
Updated: |
A VERY BRIEF HISTORY OF OPAMPS.
The first opamp to get real exposure in the UK was the National 709, a fairly primitive item which required quite complicated external compensation (which did not seem easy to get right) and was devoid of output short-circuit protection. One slip of the probe and an expensive IC was gone; I for one found this most discouraging, and gave up on the 709 at once. If you're going to quit, do it early.
To my mind, the first really practical opamp was the National 741. See the separate page devoted to it: The 741 bipolar opamp
There have however been other types that have made some impact:
The LM324.
This was one of the first quad op-amps. Its linearity was and is poor, mainly due to the rudimentary outout stage, which seemed to have poor current-sourcing capability. No noise-performance specs appear in the manufacturer's data sheet, which is not encouraging. It had some popularity because it was cheap, internally-compensated, and a quad. It may, like the 741, have some historical interest, but I will not be consider it further at present.
OPAMP THEORY.
Still in preparation. (ie I haven't started) There are lots of textbooks for the basic stuff, but naturally I want to do something different, and that takes time.
OPAMP TYPES.
Each opamp type I am familiar with has its own page, with full details and measurement results. This section just gives an overview and makes some comparisons.
Just to get the feel of things, here are the vital statistics of the most common audio opamps. Figures are all for +/-15V rails, and all specs are typical, not min or max.
TABLE 1.uA741TL072NE5534ANJM4556OPA2134
Input BJTFETBJTBJTFET
Max rails+/-15+/-18+/-20+/-18+/-18V
Output range+/-13+/-13.5+/-13+/-11*+/-17V
CM range+/-13+15 -12+/-13+/-14+/-13V
En**18488nV/rtHz
In**0.010.7**.003pA/rtHz
Ibias80.06550050+.005nA
Slew rate+/-0.5+/-13+/-9+/-3+/-20V/us
* with 150 Ohm load** No spec available | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Some opamp properties in order of: voltage noise density @ 1 kHz (Table 2) and slew rate. (Table 3)
|
|
If you want the lowest voltage noise, it has to be a bipolar input. The differences are however not huge. A modern FET-input opamp such as the OPA2134 is only 6 dB noisier than the old faithful 5532. The bipolar AD797 seems to be out on its own here.
Clearly slew rates vary more than some other parameters; a range of 100:1 is shown here. The Red Text of Shame for the 741 emphasises that this is not fast enough to give full output over the audio band.
There are of course faster ways to handle a signal, such as current-feedback architectures, but they usually fall down on linearity. In any case, a maximum slewrate greatly in excess of what is required appears to confer no benefits whatever.
OPAMP PROPERTIES: MAXIMUM SUPPLY RAILS
This parameter does not vary much across the usual opamps.
The normal practice when powering a system including TL072s and 5532s is to use +/-17V. This usually allows +20 dBu (7.75 Vrms) signals to be used internally, which is often a handy 20 dB or more above the nominal operating level, and gives a small but comforting safety margin inside the +/-18V rating of the TL072s.
OPAMP PROPERTIES: COMMON MODE RANGE
This is simply the range over which the inputs can be expected to work as proper differential inputs. It usually covers most of the range between the rail voltages, but...
Those of you viewing in colour will note that Table 1 shows the common-mode range for the TL072 in red. This is to highlight the deadly trap this IC contains for the unwary. Most opamps, when they hit their CM limits, simply show some sort of clipping. The TL072, however, when it hits its negative limit, promptly inverts its phase, so your circuit either latches up, or shows nightmare clipping behaviour with the output bouncing between the rails. The positive limit is in contrast trouble-free. See below ** for more details.
Other BiFET opamps show similar behaviour; for example the OP-42, OP-249, OP-282, AD711 and AD712. You are however unlikely to meet these parts in audio applications.
OPAMP PROPERTIES: NOISE
Table 1 shows how FET input opamps have more voltage noise but less current noise than bipolar input opamps.
OPAMP PROPERTIES: BIAS CURRENT
FET input opamps have very low bias current at room temperature; however it doubles for every 10 degree Centigrade rise. This can cause serious trouble in precision DC circuitry that must operate over a wide environmental range, but is pretty unlikely to cause trouble in audio applications.
OPAMP PROPERTIES: SLEW RATE
The TL072 slewrate is typically +/-13 V/us
The 5532 slewrate is typically +/-9 V/us. This version is internally compensated for unity-gain stability, not least because there are no spare pins for compensation when you put two opamps in an 8-pin dual package.
The single-amp version, the 5534, can afford a couple of compensation pins, and so is made to be stable only for gains of 3x or more. The basic slewrate is therefore higher at +/-13 V/us.
Compared with power-amplifier specs, which often quote 100 V/us or more, this may appear rather sluggish. In fact it is not, even +/-9 V/us being more than fast enough. Assume you are running your opamp from +/-18V rails, and that it can give a +/-17V swing on its output, For most opamps this is distinctly optimistic, but never mind. To produce a full-amplitude 20 kHz sine wave you only need 2.1 V/us, so even in the worst case there is a safety-margin of at least four times.
The 5532 is not, however, perfect. The biggest drawback is the high bias and offset currents at the inputs. This is an inevitable result of using a bipolar input stage (for low noise) without any sort of bias-cancellation circuitry. (If this was added it would compromise common-mode noise, and probably also introduce common-mode distortion, according to my conversations with IC makers) The 5532 was designed for audio work, and makes few concessions to DC accuracy.Having said that, it's not actually that bad. The offset voltage spec is 0.5 mV typ, 4 mV max, compared with 3 mV typ, 6 mV max for the popular TL072. On many occasions I have used 5532s to replace TL072s when offset voltage was a problem.
If bipolar-style bias currents cannot be accommodated, then there are plenty of FET-input opamps around. However, there is no obviously superior device that is the equivalent of the 5532. The TL072 was used for many years but its HF linearity is not first-class and distortion across the band deteriorates badly as output loading increases.
SERIES FEEDBACK.
The standard series-feedback opamp gain stage is shown at its
simplest in Fig 6. This stage gives a gain of four times (+12dB) and something similar is widely used for fader post-amplifiers in mixing consoles, and after the volume control in hifi preamps. Note that I have assumed that DC-blocking will be required for audio use; hence the input and output capacitors. These are shown as electrolytics as a large value is usually required, especially on the output.
The input capacitor Cin could be non-electrolytic, if Rin is made large enough. If Rin= 22K as shown, a 1uF capacitor would give an LF rolloff 3 dB down at 7.2 Hz, which would usually be about right- IF this is not just one stage in a chain of many. 3 dB down at 7.2 Hz means 1dB down at 14 Hz, and 0.2 dB down at 30 Hz, which is definitely inside the audio band. If many of these stages are cascaded (as in a mixer) then the 0.2 dBs-worth gradually build up and you get a very saggy bottom-end to the frequency response.
It is therefore normal practice to make each LF cutoff frequency so low that it has negligible effect in the audio band. If the LF response needs to be defined, which is a good thing, it is in done in one specific stage. Hence the 47uF input capacitor here; the -3 dB frequency is 0.2 Hz, giving an attenuation of about 0.0001 dB at 30 Hz. You can therefore cascade any number of similar stages without having to worry about the response sagging.
The output capacitor must be large enough not only to give adequate LF response, but also to avoid capacitor distortion. If the output load is 1K (and this will require something more than a TL072 to drive it with low distortion) then Cout will need to be either 47uF (-3 dB at 3.4 Hz) or 100uF (-3 dB at 1.6 Hz)
The larger value is desirable, to reduce the signal voltage across the component and so prevent capacitor distortion; this effect determines the capacitor value more than the desired frequency response does.
The series-feedback configuration has the following characteristics:
| Fig 6Series feedback opamp stage with gain of 4x. The feedback network values give suitably light loading for a TL072. If a 5532 is used instead the resistor values can be much reduced to minimise Johnson noise. (opamp2s.gif) |
| Fig 7Feedback capacitor Cnfb added, so gain is reduced to unity at DC. This is not usually necessary at low gains, but will be required with high gain to prevent input offsets being amplified to the point where they reduce the output swing. 6K8 and 47uF give a -3 dB point at 0.5 Hz, which is low enough to prevent capacitor distortion. (opamp1s.gif) |
.
SHUNT FEEDBACK.
The shunt-feedback configuration has the following characteristics:
| Fig 8The simplest possible shunt feedback opamp stage with gain of three. No input DC-blocking is shown, but there is a output blocking capacitor. (opamp5s.gif) |
| Fig 9A shunt feedback capacitor is added, and gain reduced to unity at DC. (opamp4s.gif) |
| Fig 10The bias-current causes a voltage drop in R. offset compensation. opamp3s.gif |
DRIVING HEAVY LOADS.
In some circumstances paralleled 5532 stages are appropriate, as
in Fig 12. The distortion for the single section is very high for a 5532, and it is clearly running out of current ability. Adding the second section drops the THD back to almost nothing; this can be a very simple and cost-effective way to solve the problem of driving medium loads.
O/P current boosting for headphone drivers etc.
From the above, it will be clear that the TL072 is only suitable for light loading, and even then its linearity is not good enough for the most demanding applications. The 5532 is much better, but only goes down to 500 Ohm loads.
If you need to drive a load below 500 Ohms, then IC opamps begin to
look less attractive. Opamps do exist with a greater load-driving
capability, for example the NJM4556A from JRC is capable of
driving 150 Ohms, but this device seems so far to have achieved
little market penetration. This is may be because the linearity
into more common 500 Ohm loads is distinctly inferior to the
5532.
See here for NJM4556 details: The 4556 bipolar opamp
The gains of the paralleled voltage-follower stages
are very closely equal, so there is no current-hogging, the series
output resistors allowing for any minor gain mismatches.
| Fig 12 |
 | Fig 13Paralleling four 5532 sections to drive 100 Ohms to 5 Vrms. THD plot is almost identical to two 5532 sections driving 220 Ohms (Fig 12 above) and it appears that as many can be used in parallel as required. (opamp6.gif) |
For loads beyond the capabilities of the two halves of a 5532 in parallel, it is usually more economical to adopt a hybrid circuit. These are "hybrid" circuits in that they combine IC opamps with discrete transistors; the name does not refer to thick-film construction or anything like that.
The usual procedure is to add a Class-B output stage after the opamp. A Class-A output stage is perfectly feasible if ultimate quality is required, though the power consumption is naturally much higher.
| Fig 14Combining a TL072 with a discrete Class-B output stage. The output devices are TO92 small-signal types with high beta. Forgive me if I adopt the Bob Pease style of diagram for the time being. hybcct1.gif |
The output stage in Fig 14 is the simplest possible version of Class-B, but it really works quite well. Bias is fixed by the two silicon diodes. (eg 1N4148) Inserting extra stages into feedback loops must be done with great care, to avoid adding extra phase-shifts that may cause HF instability; here there seem to be no problems, and no extra stabilising components are required.
To my eye at least, the 4K7 resistor that sets the current through the bias diodes is just begging for bootstrapping; this would minimise current variations in the diodes ought to give better sinking ability. It has to be recorded, however, that with this particular circuity it seems to make no difference at all to the linearity.
| Fig 15THD against load for TL072 with discrete Class-B output stage. No-load, 470R and 220R loading. Gain 3x, output level 7.75Vrms. hyb072.gif |
LOW-NOISE HYBRID STAGES.
Opamps are not capable of the very best noise performance,
largely due to the compromises inherent in the manufacture of an integrated circuit.
One way of getting better noise performance than an opamp can give is to use a discrete-transistor front-end coupled to an opamp that provides raw open-loop gain, and load-driving capability. The design of such hybrid stages is a non-trivial business because of the danger of HF instability.
|